ATi driver bug in Prey & Doom3 based engines

[Gstanfor]

Link
It would appear that ATi currently has a driver bug related to its new x1950 and Prey / Doom3 engined games.

Basically you end up with 2x AA instead of 4x AA on single cards, at a certain resolution but everything works properly on crossfire (which is likely what introduced the bug).

The link above has images illustrating the difference.

 

[Nightmare225]

Ouch, this is sure to spark a massive attack volley from Wreckage and retaliation from apoppin.

 

[esteban]

From the article: "No other titles exhibited this type of behavior and it didn't occur at any resolution other than 2560x1600."

While this is certainly not a good thing I don't see why it can't be a bug. I doubt many reviews use that resolution, so if this was intentional they chose a rather odd resolution to do it in.

 

[Zstream]

No other titles exhibited this type of behavior and it didn't occur at any resolution other than 2560x1600.

opps someone just said that

 

[schneiderguy]

:Q:Q:Q

I wonder if it was truly a "bug" or if they were really purposely cheating...

...considering ATI and nvidia's record of cheating in drivers, its probably a cheat

 

[josh6079]

Hmmm...odd.

Like he said:

So, the 2xAA and 4xAA scores were the same in one game; obviously some kind of bug in the driver.

He doesn't tell us what drivers he's using though.

I'll install Q4 and run some benches. I'm wondering if this is just purely an X1950 problem or if it can be replicated on the X1900's. (They do have the same cores afterall)

If not, I'll tell redbox to put up some scores with his X1950.

 

[Zenoth]

If those games were popular I might have cared a little, at least temporarily.

 

[Gstanfor]

I hope they are not cheating here, but, they already have a track record with the Doom3 engine (the Humus replacement pixel shader).

Why can't they just do what they said they would years ago and fix OpenGL? Its not like they haven't had time to since they said they would...

 

[ShadowOfMyself]

Way to go, +1 troll points with this thread

Only happens on ONE game at 2560x1600? I think that explains itself

 

[Crafty35a]

I can't believe you're actually passing this off as an intentional cheat, Gstanfor. I was pretty intrigued after reading your description, but then I clicked on the link. As others have stated, this happens in ONE resolution (and a rarely used one), in ONE game engine. Pretty clearly a bug, IMO.

 

[Zenoth]

Let a man dream I say ... let a man dream.

 

[ShadowOfMyself]

Originally posted by: Gstanfor
I hope they are not cheating here, but, they already have a track record with the Doom3 engine (the Humus replacement pixel shader).

Why can't they just do what they said they would years ago and fix OpenGL? Its not like they haven't had time to since they said they would...

Last time I checked, a x1950xtx clearly leads the 7900GTX in Quake 4, so please tell Nvidia to go fix their OpenGL:roll:

 

[smthmlk]

Originally posted by: Nightmare225
Ouch, this is sure to spark a massive attack volley from Wreckage and retaliation from apoppin.

hahahahahahahahahaha

so true!

 

[josh6079]

Whoops. Sorry, I forgot that the monitor I was going to bench-check the scores on can only go as high as 2048x1536, making it kind of difficult to check a bug that supposedly only happens at 2560x1600.

Why can't they just do what they said they would years ago and fix OpenGL? Its not like they haven't had time to since they said they would...

I wouldn't classify it as "not fixed" since any resolution besides 2560x1600 doesn't have the "bug" and therefore has shown improvements from their former, broken OpenGL. I don't see how one possible driver fluke at a rare resolution of 2560x1600 accounts for all of ATi's current OpenGL performance.

 

[redbox]

Ya it sounds like a bug to me. I don't see how some people can say that ATI drivers suck and in the same breath accuse them of enginering their drivers well enough to cheat at one resolution in one game engine. Given how bad ATI's drivers are I have no doubt that it is a driver bug. Besides very few people play at that resolution and if they do they probably have crossfire and that doesn't have this bug.

 

[Nightmare225]

Have to agree. Looks like an honest-to-god bug.

 

[Keysplayr]

Originally posted by: Gstanfor
I hope they are not cheating here, but, they already have a track record with the Doom3 engine (the Humus replacement pixel shader).

Why can't they just do what they said they would years ago and fix OpenGL? Its not like they haven't had time to since they said they would...

Didn't they? They have improved OpenGL performance quite a bit over the last 2 years. Not saying it can't be better, but it has improved. And this thing at that ONE resolution is a bug, not a cheat. Otherwise, it would happen at all resolutions. Why would a cheat be implemented on a resolution that barely anybody can play at?

 

[Gstanfor]

Well, I personally find it a little suspicious that so far the bug is limited to x1950 cards *and* occurred in the driver set that most reviewers will have tested that new card with. Who knows what other surprises these drivers may contain?

Then again crossfire is such a tangled mess anything is possible, but like said above x1900 should also suffer this problem since its the exact same core. If it doesn't I think that pretty clearly points to monkey business.

Once again though, this thread proves very nicely that if you expose an ati flaw, the lunATics laugh it off. If an nvidia flaw shows up, its the end of the world...

 

[redbox]

I would expect it to effect the x1950 more they have had driver problems since the card came out.

 

[Keysplayr]

Originally posted by: Gstanfor
Well, I personally find it a little suspicious that so far the bug is limited to x1950 cards *and* occurred in the driver set that most reviewers will have tested that new card with. Who knows what other surprises these drivers may contain?

Then again crossfire is such a tangled mess anything is possible, but like said above x1900 should also suffer this problem since its the exact same core. If it doesn't I think that pretty clearly points to monkey business.

Once again though, this thread proves very nicely that if you expose an ati flaw, the lunATics laugh it off. If an nvidia flaw shows up, its the end of the world...

Legendary moronic proportions. I'm an nvidia fan, and I can clearly see this is a bug. How many hours per day do you stretch your arms? They got to be 8 1/2 to 9 feet by now. The better to "reach" with.

 

[darkswordsman17]

Originally posted by: keysplayr2003

Originally posted by: Gstanfor
Well, I personally find it a little suspicious that so far the bug is limited to x1950 cards *and* occurred in the driver set that most reviewers will have tested that new card with. Who knows what other surprises these drivers may contain?

Then again crossfire is such a tangled mess anything is possible, but like said above x1900 should also suffer this problem since its the exact same core. If it doesn't I think that pretty clearly points to monkey business.

Once again though, this thread proves very nicely that if you expose an ati flaw, the lunATics laugh it off. If an nvidia flaw shows up, its the end of the world...

Legendary moronic proportions. I'm an nvidia fan, and I can clearly see this is a bug. How many hours per day do you stretch your arms? They got to be 8 1/2 to 9 feet by now. The better to "reach" with.

No kidding. You need therapy or something as its clear your vision of reality has been skewed by online fanboyism.

 

[Gstanfor]

Remember the 8500 and QUAK? another OpenGL "driver bug"...

One can only wonder what Carmack must make of all these bugs that somehow manage to involve his game engines almost everytime ATi releases a new card...

 

[cmdrdredd]

Originally posted by: Gstanfor
Well, I personally find it a little suspicious that so far the bug is limited to x1950 cards *and* occurred in the driver set that most reviewers will have tested that new card with. Who knows what other surprises these drivers may contain?

Then again crossfire is such a tangled mess anything is possible, but like said above x1900 should also suffer this problem since its the exact same core. If it doesn't I think that pretty clearly points to monkey business.

Once again though, this thread proves very nicely that if you expose an ati flaw, the lunATics laugh it off. If an nvidia flaw shows up, its the end of the world...

No what this shows is people who are clearely Nvidia fanboys like to find any little thing to hide the fact that Nvidia has and still is cheating in DX9 and NOT adhering fully to the specifications of the API.

 

[40sTheme]

Originally posted by: keysplayr2003

Originally posted by: Gstanfor
Well, I personally find it a little suspicious that so far the bug is limited to x1950 cards *and* occurred in the driver set that most reviewers will have tested that new card with. Who knows what other surprises these drivers may contain?

Then again crossfire is such a tangled mess anything is possible, but like said above x1900 should also suffer this problem since its the exact same core. If it doesn't I think that pretty clearly points to monkey business.

Once again though, this thread proves very nicely that if you expose an ati flaw, the lunATics laugh it off. If an nvidia flaw shows up, its the end of the world...

Legendary moronic proportions. I'm an nvidia fan, and I can clearly see this is a bug. How many hours per day do you stretch your arms? They got to be 8 1/2 to 9 feet by now. The better to "reach" with.

Haha, that last comment made me laugh really hard.
Once again though, this Gstanfor proves very nicely that if you expose and Nvidia flaw, he laughs it off. If an ATi flaw shows up, its the end of the world...

 

[thilanliyan]

Originally posted by: Gstanfor
Well, I personally find it a little suspicious that so far the bug is limited to x1950 cards *and* occurred in the driver set that most reviewers will have tested that new card with. Who knows what other surprises these drivers may contain?

Then again crossfire is such a tangled mess anything is possible, but like said above x1900 should also suffer this problem since its the exact same core. If it doesn't I think that pretty clearly points to monkey business.

If you were comparing maybe an X1900XT 256mb vs. 512 mb and the problem ocurred on only one card then maybe you could call it a cheat because the cards are NEARLY identical. The X1950 doesn't even use the same type of memory which AFAIK is involved in the processing of 3D images. Maybe it's the way the driver interacts with the GDDR4 at that resolution??? You're REALLY reaching with this.